Sample Size of One
The dangerous psychological trap of substituting personal anecdotes for financial modeling.
When Peter Lynch was managing the legendary Magellan Fund, he famously bought Dunkin’ Donuts.
He didn’t use a proprietary algorithmic model to find it. He didn’t read a 400-page academic white paper on the future of carbohydrates. He just noticed the long lines, loved the coffee, and bought the stock.
Wall Street laughed at him (see video below). Retail investors thought it was a joke. A coffee and donut shop didn’t sound smart enough to be a serious financial thesis.
We have a deep-seated psychological need for our investments to stroke our intellectual egos.
But here is the “A-ha” moment that everyone forgets about that story: Lynch didn’t buy Dunkin’ Donuts just because the coffee was good.
He bought it because, after noticing the lines, he looked at the books and saw a hyper-resilient, scalable business with exceptional unit economics.
Unfortunately, modern investors only remember the first half of the story. They weaponized Lynch’s famous “buy what you know” mantra to justify lazy investing.
This brings us to a strange, persistent psychological trap in the market, that I see is going on right now: The more accessible and easy-to-understand a product is, the lazier our financial analysis becomes.
The “Sample Size of One” Trap
Whenever I discuss a consumer app like Duolingo, my feed instantly fills up with anecdotal analysis:
“I downloaded the free version, tried it for an evening, didn’t like the gamification, and deleted it. The moat is dead.”
Or my personal favorite:
“My father used it for 5 years and did not learn a single piece of Spanish.”
It is absolutely fascinating that this collective failure of logic only happens with consumer-facing products. Because everyone knows how to swipe a screen or tap a green owl, they subconsciously believe they understand the global addressable market.
They are confusing a user with a cohort.
Your father? He is a data point of one.
Duolingo has nearly 100 million monthly active users. If their engineering team tweaks the gamification algorithm and increases free-to-paid conversion by just 0.5%, your dad won’t notice, and your personal user experience won’t change. But that tiny shift generates tens of millions in high-margin recurring revenue.
When a product is simple, we substitute actual financial modeling for a Yelp review. We mistake our own reflection for the macroeconomic environment.
The Complexity Bias & The “Whisper Stock”
But here is where the hypocrisy gets truly entertaining.
This anecdotal arrogance completely vanishes the second things get complicated.
You rarely see this with complex, established tech. Nobody says:
“I tried an $AMD EPYC server processor for a week in my basement, didn’t like the architecture latency, so I’m shorting the stock.”
With AMD, people are forced to actually read the fundamentals. Nobody drives past an $IREN facility, decides the liquid cooling hum is “bad vibes,” and assumes the stock is doomed.
With IREN, investors actually build spreadsheets to model out power pipelines and cost-per-megawatt.
But what happens when the tech is so complex that there are no fundamentals to model yet?
We surrender to the hype train.
We suffer from “Complexity Bias”; the tendency to give undue credit to things we don’t understand. When faced with the complex AI, space tech, or ESG waves, people happily abandon logic. They assume that if they can’t understand the tech, the people building it must be geniuses, and therefore, it’s a guaranteed winner.
Peter Lynch used to call these “Whisper Stocks.” These are the companies with hypnotic, world-changing stories, complex sci-fi pitch decks, and absolutely zero revenue. You hear about them at a dinner party: “I hear they’re going to revolutionize liquid-cooled AI infrastructure...”
This strategy is catastrophic when used recklessly. Investors blindly buy into the peak of the hype cycle, mistaking incomprehensibility for inevitability. We overthink the simple, profitable apps because we understand them, and we blindly buy the unprofitable tech infrastructure because we don’t.
Building an Atomic Moat Around Your Mind
So, how do we stop being anecdotal tourists and start investing with discipline? We have to invert our natural psychological defaults.
1. For the “Simple” Stuff: Track the Cohort, Not the Cousin.
You don’t have to love the Duolingo owl. You don’t have to enjoy gamified learning. But to find a true atomic moat, you must respect the aggregate data.
Look at DAU/MAU Growth: Are people showing up globally?
Look at Free-to-Paid Conversion: Is the engine actually working?
Look at Churn: Are the paid users sticking around?
Right now, for companies like Duolingo, those fundamentals look incredibly strong, completely regardless of whether you enjoyed your 5-minute French lesson.
2. For the “Complex” Stuff: Demand the Receipt.
Stop letting a sci-fi story override a lack of fundamentals. Whisper stocks are great for cocktail parties and terrible for portfolios. Look for actual revenue, signed enterprise contracts, and a realistic path to profitability. Do not invest recklessly just because a slide deck uses the word “quantum” or “LLM.”
Your portfolio doesn’t care if you enjoyed your Spanish lesson, and it doesn’t care how cool a liquid-cooled GPU cluster sounds.
Stop treating your investments like a Yelp review. Follow the data.
Rob H.
Disclaimer: The content of Atomic Moat is for educational and entertainment purposes only and does not constitute financial, investment, or legal advice. I am not a financial advisor, and these are not recommendations to buy or sell any security.
Risk Warning: Investing in equities, especially in the technology sector, involves a high degree of risk. The market is irrational, and prices can fluctuate wildly. Please do your own due diligence (DYODD) and consult with a certified professional before making any investment decisions.